SBC Digital Sportsbook: Artificial intelligence panel says the technology is powerful but needs guardrails
SBC Digital Sportsbook: Artificial intelligence panel says the technology is powerful but needs guardrails
By Rege Behe
Moderator James Ross opened Wednesday’s SBC Digital –Sportsbook panel on artificial intelligence by paraphrasing, of all things, Spider-Man.
“With great power comes great responsibility,” said Ross, a multimedia editor for SBC Media. “With innovation comes great responsibility. The risk of AI bias, regulatory scrutiny, and liabilities cannot be ignored. How do we ensure that AI serves as a force of progress rather than a sort of pitfall?”
According to the participating panel, AI must be deployed with caution.
“I think there’s great, great potential, and it’s evolving quickly,” said Bettson Group Engineering Team Lead Camilo Lucena Herradez. “At the same time, I think it’s not only in sport, but it’s going to [be used] slowly in the personalization part. So, with more data collected from users, it’s slowly going to adapt and personalize the offerings you can see in sports and how they interact with the user.”
“AI is not technology, it’s a tool,” said George Wawer, Win2Day Managing Director. “It’s going to be everywhere. It’s like a commodity. […] Right now, it’s one product for many customers. We do believe that we have to customize it, to an extent, so it’s one product made for you.”
But, Wawer added, “AI is only as good as the data we feed it.”
David Jacquet, Tzeract Managing Director of AI Trading, said that the information input must be correct, notably for sporting events, because “there are more odds combinations on a game than there are atoms in the sun.”
“So, to predict everything in a good way is very difficult,” Jacquet added.
The data may include historical prices, event data, event feed data and countless other types of information.
“We’re trying to predict a huge, huge set of output with a limited set of input,” Jacquet said. “And if you have the wrong human designing the actual solution, the bias problem can be astronomical because the models can be over-trained, they can solve the wrong problems.
“For a bookmaker, it’s a little bit of a defensive activity. In my opinion, you’re not valued on your best, but at your worst. The biggest thing for a bookmaker is to avoid disastrous mistakes. There are always bad prices on any bookmaker site, and the thing you really need to do is not lose too much money on that.”
Herradez said there are a couple of ways sportsbooks can protect themselves, notably through data filtering.
“Be mindful of the date you are using,” Herradez said. “Also, review the data constantly. […]You need to validate this data. You can get, for example, data from 2020 and go over your models and try to predict what is going to happen the next year and use human resources, as well, to validate these inputs.
Wawer said it is important to factor in and input all considerations. An example would be trying to predict an outcome for a team’s next season at the end of the previous season.
“Team composition has changed,” Wawer said, noting that factors such as a player moving from one team to another are often not considered and other considerations, such as team makeup, are overlooked. “Not all the data points [are considered] and therefore the data is wrong.
“I think the main principle must be that there’s always a human in the end, pushing the buttons, making the main decisions. What we try to do is either limit it to very low risk or have a human overwatch.”
Verticals:
Sectors:
Topics:
